DNA methylation and transcriptional gene silencing (TGS) play important roles in genome organization and stability, genomic imprinting, X chromosome inactivation and other developmental processes in eukaryotes. Little is known about the initial trigger for DNA methylation that is important for stable TGS, and in particular, the cellular mechanisms for the active suppression of TGS are not understood. We have developed a unique TGS system in the model organism Arabidopsis thaliana. In this system, an active transgene and a homologous endogenous gene become silenced when cellular ROS (repressor or silencing) factors are mutated. We have shown that ROS1 encodes a DNA glycosylase/lyase that prevents the hypermethylation and TGS of the homologous genes by active DNA demethylation via a base excision repair mechanism. We propose here to characterize the putative DNA demethylation activity of ROS1, to clone the ROS2 and ROS3 loci, to identify ROS1-interacting proteins, and to isolate and clone ros1 suppressor mutations. The mechanisms of active DNA demethylation have been very controversial, in part due to a critical lack of genetic evidence for the in vivo function of putative DNA demethylases. Our combined genetic and biochemical analysis of the ROS genes and proteins promises to provide unequivocal evidence that a base excision repair DNA glycosylase is an active demethylase, and will lead to the isolation of novel components of cellular pathways for the suppression of TGS.